A patrol robot equipment identification method and system

Abstract

The present invention relates to a patrol robot device identification method and system, wherein the method performs image preprocessing on initial image information to obtain background information and device information to be tested; classifies the background information and the device information to be tested respectively based on a machine learning algorithm to obtain background classification information and device classification information; fuses the background classification information and the device classification information to obtain fusion information; matches the fusion information based on a preset device library to obtain a device identification result. The present invention automatically classifies and identifies images through a machine learning algorithm, thereby improving the automation level of device identification of the patrol robot.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent inspection of electric power systems, and in particular to an inspection robot equipment identification method and system. Background Art

[0002] Ensuring the safe, stable and reliable operation of UHV substations is currently the top priority, especially for UHV substations. The 1000kV GIS equipment has a compact structure and irregular inspection paths. Conventional outdoor intelligent inspection robots cannot automatically inspect the equipment in the equipment area. At the same time, the existing in-station inspection robots and video surveillance have single functions and do not have image analysis and intelligent detection functions. Operation and maintenance personnel are required to manually perform abnormal inspections and defect detection, which cannot support the needs of lean operation and maintenance control of substations, cannot exert integrated and coordinated forces, and are difficult to fundamentally reduce the workload of operation and maintenance personnel. Summary of the invention

[0003] In order to overcome the deficiencies of the prior art, an object of the present invention is to provide a method and system for equipment identification of an inspection robot, which can improve the degree of automation of equipment identification of the inspection robot.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] A method for identifying inspection robot equipment, comprising:

[0006] Obtain the initial image information of the device under test;

[0007] Perform image preprocessing according to the initial image information to obtain background information and information of the device to be tested;

[0008] Based on a machine learning algorithm, the background information and the device information to be tested are classified respectively to obtain background classification information and device classification information;

[0009] fusing the background classification information and the device classification information to obtain fused information;

[0010] The fusion information is matched based on a preset device library to obtain a device identification result.

[0011] Preferably, performing image preprocessing according to the initial image information to obtain background information and device under test information includes:

[0012] Extract pixels according to the initial image information to obtain extracted information and unextracted information;

[0013] Pixel merging is performed on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

[0014] Preferably, after performing image preprocessing according to the initial image information to obtain background information and information of the device to be tested, the method further includes:

[0015] The background information and the device information to be tested are subjected to image binarization processing respectively to obtain the background information after image refinement and the device information to be tested after image refinement.

[0016] Preferably, the background information and the device information to be tested are classified based on a machine learning algorithm to obtain background classification information and device classification information, including:

[0017] Using the trained image recognition model to respectively identify the background information and the device information to be tested, to obtain background recognition information and device recognition information;

[0018] The training method of the image recognition model is:

[0019] Obtain training segmentation image data;

[0020] Annotating the training segmented image data to obtain annotated image data;

[0021] Determining a loss function based on the labeled image data and the training segmented image data;

[0022] Minimizing the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model;

[0023] Determine whether the number of training times reaches the preset number of training times. If not, update the weight of the image recognition model according to the update gradient, and input the training segmented image data into the updated image recognition model for segmentation training; if the preset number of training times is reached, output the trained image recognition model.

[0024] Preferably, the fusing the background classification information and the device classification information to obtain fused information includes:

[0025] Constructing a framework according to the background classification information and the device classification information to obtain a recognition framework;

[0026] The discrimination framework is parsed based on the DS evidence reasoning method to obtain the fusion information.

[0027] A patrol robot equipment identification system, comprising:

[0028] An acquisition module, used to acquire initial image information of the device under test;

[0029] A preprocessing module, used to perform image preprocessing according to the initial image information to obtain background information and information of the device to be tested;

[0030] A classification module, used to classify the background information and the device information under test respectively based on a machine learning algorithm to obtain background classification information and device classification information;

[0031] A fusion module, used for fusing the background classification information and the device classification information to obtain fusion information;

[0032] The matching module is used to match the fusion information based on a preset device library to obtain a device identification result.

[0033] Preferably, the preprocessing module comprises:

[0034] An extraction unit, used to extract pixels according to the initial image information to obtain extracted information and unextracted information;

[0035] The merging unit is used to perform pixel merging on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

[0036] Preferably, it also includes:

[0037] The refinement module is used to perform image binarization processing on the background information and the device information to be tested respectively, so as to obtain the background information after image refinement and the device information to be tested after image refinement.

[0038] Preferably, the classification module comprises:

[0039] An identification unit, used to identify the background information and the device information under test respectively by using a trained image recognition model to obtain background identification information and device identification information;

[0040] The identification unit specifically includes:

[0041] An acquisition subunit, used for acquiring training segmentation image data;

[0042] A labeling subunit, used for labeling the training segmented image data to obtain labeled image data;

[0043] A determination subunit, configured to determine a loss function based on the labeled image data and the training segmented image data;

[0044] An updating subunit, used to minimize the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model;

[0045] The judgment subunit is used to judge whether the number of training times reaches the preset number of training times. If not, the weight of the image recognition model is updated according to the update gradient, and the training segmented image data is input into the updated image recognition model for segmentation training; if the preset number of training times is reached, the trained image recognition model is output.

[0046] Preferably, the fusion module includes:

[0047] A construction unit, configured to construct a framework according to the background classification information and the device classification information to obtain a recognition framework;

[0048] The parsing unit is used to parse the discrimination framework based on the DS evidence reasoning method to obtain the fusion information.

[0049] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0050] The present invention provides a method and system for device identification of a patrol robot. The method performs image preprocessing on initial image information to obtain background information and device information to be tested; classifies the background information and the device information to be tested respectively based on a machine learning algorithm to obtain background classification information and device classification information; fuses the background classification information and the device classification information to obtain fusion information; matches the fusion information based on a preset device library to obtain a device identification result. The present invention automatically classifies and identifies images through a machine learning algorithm, thereby improving the automation level of device identification of the patrol robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0052] Figure 1 A flow chart of a method for identifying an inspection robot device in an embodiment of the present invention;

[0053] Figure 2 Module connection diagram of the inspection robot equipment identification system in the embodiment provided by the present invention DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] The purpose of the present invention is to provide a patrol robot equipment identification method and system, which can improve the automation level of equipment identification of the patrol robot.

[0056] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] Figure 1 The flow chart of the inspection robot device identification method in the embodiment provided by the present invention is as follows: Figure 1 As shown, a patrol robot device identification method in this embodiment includes:

[0058] Step 100: Acquire initial image information of the device under test;

[0059] Step 200: performing image preprocessing according to the initial image information to obtain background information and information of the device to be tested;

[0060] Step 300: Classify the background information and the device information to be tested respectively based on a machine learning algorithm to obtain background classification information and device classification information;

[0061] Step 400: Fusing the background classification information and the device classification information to obtain fused information;

[0062] Step 500: Match the fusion information based on a preset device library to obtain a device identification result.

[0063] Preferably, performing image preprocessing according to the initial image information to obtain background information and device under test information includes:

[0064] Extract pixels according to the initial image information to obtain extracted information and unextracted information;

[0065] Pixel merging is performed on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

[0066] Specifically, when using this technology to identify an image, the image to be identified can be input into the computer system first, and then the computer will automatically match the image, extract the characteristics of the image, and perform specific classification. In the process of intelligent image recognition, the preprocessing link is image input, which is a relatively critical link and is directly related to the image recognition effect. The collected image is input into the computer, and at this time the recognition system will separate the background from the image according to the characteristics of the image itself.

[0067] Preferably, after performing image preprocessing according to the initial image information to obtain background information and information of the device to be tested, the method further includes:

[0068] The background information and the device information to be tested are subjected to image binarization processing respectively to obtain the background information after image refinement and the device information to be tested after image refinement.

[0069] Optionally, the binarization of the image is refined as a whole, thereby ensuring the speed and efficiency of the system's subsequent image processing, thereby ensuring the authenticity of the image to the greatest extent, reducing the false components therein, and completing the image recognition with the help of relevant numerical values.

[0070] Preferably, the background information and the device information to be tested are classified based on a machine learning algorithm to obtain background classification information and device classification information, including:

[0071] Using the trained image recognition model to respectively identify the background information and the device information to be tested, to obtain background recognition information and device recognition information;

[0072] The training method of the image recognition model is:

[0073] Obtain training segmentation image data;

[0074] Annotating the training segmented image data to obtain annotated image data;

[0075] Determining a loss function based on the labeled image data and the training segmented image data;

[0076] Minimizing the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model;

[0077] Determine whether the number of training times reaches the preset number of training times. If not, update the weight of the image recognition model according to the update gradient, and input the training segmented image data into the updated image recognition model for segmentation training; if the preset number of training times is reached, output the trained image recognition model.

[0078] Specifically, this embodiment can use a target detection algorithm to train an image recognition model, such as the MaskR-CNN algorithm, the FasterR-CNN algorithm, or the FastR-CNN algorithm, etc. In addition, the recognition accuracy of the visible light image recognition model must exceed a preset threshold value before the trained image recognition model is considered qualified.

[0079] Optionally, the threshold value is 0.9.

[0080] Preferably, the fusing the background classification information and the device classification information to obtain fused information includes:

[0081] Constructing a framework according to the background classification information and the device classification information to obtain a recognition framework;

[0082] The discrimination framework is parsed based on the DS evidence reasoning method to obtain the fusion information.

[0083] As an optional implementation method, computer intelligent image recognition belongs to the category of multi-source information processing. In the processing of such information, data fusion is a brand-new technology, and DS evidence reasoning is one of the main research methods. It is based on the non-empty set θ, which is also called the discriminant framework. This framework can describe the set of all elements that constitute the hypothesis space.

[0084] Figure 2 The module connection diagram of the inspection robot equipment identification system in the embodiment provided by the present invention is as follows: Figure 2 As shown, a patrol robot equipment identification system in this embodiment includes:

[0085] An acquisition module, used to acquire initial image information of the device under test;

[0086] A preprocessing module, used to perform image preprocessing according to the initial image information to obtain background information and information of the device to be tested;

[0087] A classification module, used to classify the background information and the device information under test respectively based on a machine learning algorithm to obtain background classification information and device classification information;

[0088] A fusion module, used for fusing the background classification information and the device classification information to obtain fusion information;

[0089] The matching module is used to match the fusion information based on a preset device library to obtain a device identification result.

[0090] Preferably, the preprocessing module comprises:

[0091] An extraction unit, used to extract pixels according to the initial image information to obtain extracted information and unextracted information;

[0092] The merging unit is used to perform pixel merging on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

[0093] Preferably, it also includes:

[0094] The refinement module is used to perform image binarization processing on the background information and the device information to be tested respectively, so as to obtain the background information after image refinement and the device information to be tested after image refinement.

[0095] Preferably, the classification module comprises:

[0096] An identification unit, used to identify the background information and the device information under test respectively by using a trained image recognition model to obtain background identification information and device identification information;

[0097] The identification unit specifically includes:

[0098] An acquisition subunit, used for acquiring training segmentation image data;

[0099] A labeling subunit, used for labeling the training segmented image data to obtain labeled image data;

[0100] A determination subunit, configured to determine a loss function based on the labeled image data and the training segmented image data;

[0101] An updating subunit, used to minimize the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model;

[0102] The judgment subunit is used to judge whether the number of training times reaches the preset number of training times. If not, the weight of the image recognition model is updated according to the update gradient, and the training segmented image data is input into the updated image recognition model for segmentation training; if the preset number of training times is reached, the trained image recognition model is output.

[0103] Preferably, the fusion module includes:

[0104] A construction unit, configured to construct a framework according to the background classification information and the device classification information to obtain a recognition framework;

[0105] The parsing unit is used to parse the discrimination framework based on the DS evidence reasoning method to obtain the fusion information.

[0106] The beneficial effects of the present invention are as follows:

[0107] (1) The present invention automatically classifies and identifies images through a machine learning algorithm, thereby improving the automation level of equipment identification of the inspection robot.

[0108] (2) The present invention refines the image as a whole by binarizing it, thereby improving the speed and efficiency of the system's subsequent image processing, thereby ensuring the authenticity of the image to the greatest extent and reducing the false components therein.

[0109] (3) The present invention adopts a fusion algorithm to make the recognition result more reliable.

[0110] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0111] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for identifying inspection robot equipment, characterized in that: include: Obtain the initial image information of the device under test; Perform image preprocessing according to the initial image information to obtain background information and information of the device to be tested; Based on a machine learning algorithm, the background information and the device information to be tested are classified respectively to obtain background classification information and device classification information; The background classification information and the device classification information are integrated to obtain integrated information, specifically including: constructing a framework according to the background classification information and the device classification information to obtain a recognition framework; The discrimination framework is parsed based on the DS evidence reasoning method to obtain the fusion information; The fusion information is matched based on a preset device library to obtain a device identification result.

2. The inspection robot equipment identification method according to claim 1, characterized in that: The performing image preprocessing according to the initial image information to obtain background information and information of the device to be tested includes: Extract pixels according to the initial image information to obtain extracted information and unextracted information; Pixel merging is performed on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

3. The inspection robot equipment identification method according to claim 1, characterized in that: After performing image preprocessing according to the initial image information to obtain background information and information of the device to be tested, the method further includes: The background information and the device information to be tested are subjected to image binarization processing respectively to obtain the background information after image refinement and the device information to be tested after image refinement.

4. The inspection robot equipment identification method according to claim 1, characterized in that: Based on a machine learning algorithm, the background information and the device information to be tested are classified respectively to obtain background classification information and device classification information, including: Using the trained image recognition model to respectively identify the background information and the device information to be tested, to obtain background recognition information and device recognition information; The training method of the image recognition model is: Obtain training segmentation image data; Annotating the training segmented image data to obtain annotated image data; Determining a loss function based on the labeled image data and the training segmented image data; Minimizing the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model; Determine whether the number of training times reaches the preset number of training times. If not, update the weight of the image recognition model according to the update gradient, and input the training segmented image data into the updated image recognition model for segmentation training; if the preset number of training times is reached, output the trained image recognition model.

5. A patrol robot equipment identification system, characterized in that: include: An acquisition module, used to acquire initial image information of the device under test; A preprocessing module, used to perform image preprocessing according to the initial image information to obtain background information and information of the device to be tested; A classification module, used to classify the background information and the device information under test respectively based on a machine learning algorithm to obtain background classification information and device classification information; A fusion module, used for fusing the background classification information and the device classification information; The fusion module includes: A construction unit, configured to construct a framework according to the background classification information and the device classification information to obtain a recognition framework; A parsing unit, used for parsing the discrimination framework based on the DS evidence reasoning method to obtain fusion information; The matching module is used to match the fusion information based on a preset device library to obtain a device identification result.

6. The inspection robot equipment identification system according to claim 5, characterized in that: The preprocessing module comprises: An extraction unit, used to extract pixels according to the initial image information to obtain extracted information and unextracted information; The merging unit is used to perform pixel merging on the extracted information and the unextracted information respectively to obtain the background information and the device under test information.

7. The inspection robot equipment identification system according to claim 5, characterized in that: Also includes: The refinement module is used to perform image binarization processing on the background information and the device information to be tested respectively, so as to obtain the background information after image refinement and the device information to be tested after image refinement.

8. The inspection robot equipment identification system according to claim 5, characterized in that: The classification module comprises: An identification unit, used to identify the background information and the device information under test respectively by using a trained image recognition model to obtain background identification information and device identification information; The identification unit specifically includes: An acquisition subunit, used for acquiring training segmentation image data; A labeling subunit, used for labeling the training segmented image data to obtain labeled image data; A determination subunit, configured to determine a loss function based on the labeled image data and the training segmented image data; An updating subunit, used to minimize the value of the loss function using a gradient descent optimizer to obtain an updated gradient of the image recognition model; The judgment subunit is used to judge whether the number of training times reaches the preset number of training times. If not, the weight of the image recognition model is updated according to the update gradient, and the training segmented image data is input into the updated image recognition model for segmentation training; if the preset number of training times is reached, the trained image recognition model is output.

Images